The free tier is modest on purpose. If you want to use IsraelGPT from a coding agent or run something bigger, apply for the Pro API: you say what it is for, a person reads it, and sets limits that fit.
More of everything the free tier limits, on every key of your account, and room for the long prompts agents send:
| Requests a minute | Requests a day | Tokens a day | Largest prompt | Longest reply | Keys | |
|---|---|---|---|---|---|---|
| Free | 10 | 200 | 400,000 | 12,000 tokens | 2,000 tokens | 5 |
| Starter | 30 | 500 | 1,000,000 | 32,000 tokens | 4,000 tokens | 5 |
| Standard | 60 | 1,500 | 4,000,000 | 64,000 tokens | 8,000 tokens | 10 |
| Heavy | 120 | 4,000 | 12,000,000 | 128,000 tokens | 16,000 tokens | 20 |
These are the usual starting points. The person who reviews your application can set other numbers that suit what you describe, so ask for what you actually need. Free counts requests per key (and per account); a Pro plan counts them per account, across all your keys, so more keys don’t mean more requests. Tokens are what the model counts, prompt and reply together, per day (UTC).
The Pro API is the same API: OpenAI-compatible and the native endpoint, the same keys, the same models. Only the limits change.
It isn’t for reselling access, sharing keys between unrelated people, or building a service whose users you can’t account for. Those are against the Terms of Service, and a plan can be changed or ended if it’s misused.
What we ask:
| Question | Why |
|---|---|
| Project name | To tell your application from the others. |
| What is it for? (80 to 2,000 characters) | What you’re building, who uses it, and why the free limits aren’t enough. The most useful answer is a specific one. |
| Tools you’ll use | Cline, Roo Code, Continue, Aider, Open WebUI, your own app, a bot, or something else. |
| Requests a day, and the most in one minute | Sets the request limits. Be honest and a little generous; you can ask again if you outgrow it. |
| Typical prompt size | Sets the prompt limit: short chats, up to 32k, up to 64k, or 100k+ tokens for an agent reading a codebase. |
| Who will use it | Only you or your team, a few other people, or many users. |
| A link (optional) | A page or repository about the project. https only. |
| The API rules | That you’ll use it for what you described, and won’t resell it or share your keys. |
Your answers are kept encrypted, are read only by the people who review applications (the owner, and moderators allowed to), and go when you delete your account. They’re checked like anything else you write on the site.
| Decision | What happens |
|---|---|
| Approved | You get the limits, an email with them, and an end date if the plan has one. Apply again before it ends to keep it. |
| Not approved | You get an email with a note, usually about what to add. You can apply again after 14 days, or sooner if the reviewer says so. Your keys keep working on the free limits. |
| Withdrawn | You can take back an application that’s waiting, from the dashboard, and send a new one. |
| Changed or ended later | A plan can be changed, or ended with a reason if it’s misused. Your keys aren’t revoked: they go back to the free limits. |
One application can wait at a time. Questions: support@israelgpt.site.
429 and Retry-After. A program that ignores them is paused for longer.